It may be seen as an accident of nature, and atypical among mammals, that human vision is so advanced. As large-brained, large-eyed primates, humans have come to rely strongly upon vision, as it carries with many advantages, such as perceiving individuals and events from a distance. Primates in general take vision a step further than other mammals, using vision to observe others, evaluate social relationships and situations, select mates, and predict behavior. Vision is thus at the center of many aspects of human cognition related not only to our navigation and object recognition, but also to our interaction with others. Thus understanding vision is critical for understanding human cognition. Our research program has its roots in the study of visual perception, from its barest raw signals coming out of the retina to its most complex interpretations in the prefrontal cortex. We employ a range of experimental research tools to ask questions such as, how can the brain impose three-dimensional structure on its internal representation of the world, if its retinal images are inherently two-dimensional? This question, familiar to vision scientists, may seem nonsensical to those who have not yet pondered the fact that there is a problem there. When one looks at the 3-D world, or a 2-D photograph for that matter, the brain takes patterns of light and dark, color and texture, and composes the impression of three-dimensional space. These mechanisms are so hard wired, that it is impossible to avoid seeing three-dimensional structure if the appropriate cues are there. There are countless such cues, but to name a few, consider the tricks an artist uses to capture depth: perspective, shadows, texture gradients, foreshortening, and occlusion. These and many others are automatically absorbed and unconsciously interpreted be the depth, avoiding any moment in which the world might appear as it is on the retina: flat. We exploit particular cues for studying depth by stripping stimuli down to their bare minimum and leaving one cue, say 2-D shape, to define the three-dimensional structure. Within a range of parameters, virtually every such cue lends itself to perceptual ambiguity, where different people might see the same 2-D structure adopting different 3-D configurations. In the right regime, the number of potential interpretations becomes exactly two. And faced with this situation, the brain does something rather unexpected; it continually changes its mind, alternating subjectively between the two possible interpretations every few seconds. Such stimuli, generally termed bistable, are useful tools for neuroscientists trying to understand the principles of visual perception. For example, one is able to ask questions such as, if the stimulus on the retina is constant, but the perceptual interpretation spontaneously changes, where in the brain do neural responses reflect the unchanging stimulus and where to they instead reflect the changing percept? This question is at the heart of much of our research. We have recently completed three studies related to the mechanisms underlying visual perception, with two others currently underway. One recently published study used a famous visual illusion, called a Kanizsa figure (Cox et al., Proc Natl Acad Sci (2013)), in which four pac man shapes face one another in a square array. This configuration gives rise to the illusion of a surface that appears to float in front of the page, or in our case in front of the computer monitor screen. We measured neural responses in area V4 while macaques viewed this stimulus under a number of configurations. A subset of these configurations gave rise to the perception of a floating illusory surface, as described, whereas the rest were control stimuli that did not. Importantly, the basic structure of the control stimuli was in nearly all ways similar to the main stimuli, except that the pac man elements were facing a different direction. Thus we wanted to see whether neurons in area V4 responded in an enhanced manner during the illusion, as this may provide clues to how the brain infers surfaces from visual cues more generally. We found that V4 neurons indeed responded much more vigorously during perception of the illusory surface than in any of the control conditions. Moreover, the firing took a different form, with large oscillations during floating surface perception that were not present when the image appeared only flat. Interestingly, the role of a given neuron in this illusion was highly sensitive to the position of its receptive field, or the region of visual space that it monitored. Small shifts of the stimulus relative to this receptive field changed the nature of the perceptual neural modulation. Finding these results in V4, including both the perception-related modulation and the sensitivity to spatial position, adds new and potentially important evidence that this important part of the primate visual cortex is involved in the subjective construction of the third dimension of visual space. In another study (Murphy et al., Front Psychol (2014)), we investigated one salient feature of bistable visual perception that we discovered more than a decade ago (Leopold et al., Nature Neuroscience (2002)), which is that although the probability of seeing bistable stimulus a particular way on a given presentation may be likened to flipping a coin, things are very different when the stimulus is shown repeatedly. Namely, repeated presentation of the same bistable stimulus leads to a stable perception of one or the other interpretation that can last for several minutes, rather than just a few seconds. In our recent study, we demonstrated that this effect can last much longer yet, and is fundamentally driven by a new type of perceptual memory whose nature is only partially understood and whose underlying neural mechanisms are virtually unknown. This study provides the basis for designing fMRI and electrophysiological experiments in the future to investigate the neural mechanisms. In a final study, we exploited the phenomenon of blindsight to study the information carried by visual pathways under conditions in which V1 is damaged. In blindsight, there is a fascinating dissociation between the subjective perception of a person, who claims he or she cannot see anything, and their preserved capacity to use vision to guide certain aspects of their behavior, such as pointing to or reaching for an object. Our most recent study is a follow-up in a series of studies related to blindsight, which have been reported in previous years. In the most recent study, we found that the field potentials in area V4 were affected by a V1 lesion, to be expected. However, in contrast to the spiking of neurons carrying visual information about a stimulus, which are sharply diminished, the amplitude of stimulus-induced field potentials increased following the V1 lesion. These results suggest that the residual pathways influencing cortical responses following V1 lesions affect the circuitry in a manner that is very distinct from the normal feedforward pathways. While speculative, it is intriguing to think that this different mode of activity relates directly to the dissociation between perception and action, for which blindsight is famous.